metadata
metrics:
- accuracy
pipeline_tag: audio-classification
tags:
- music
This model is finetuned version of ntu-spml/distilhubert for music genre classification.
This model was finetuned to distinguish between 18 music genres including:
- Blues
- Classical music
- Country music
- Drum & Bass
- Dubstep
- Folk
- Future Bass
- Hardstyle
- House
- Jazz
- Latin music
- Metal
- Pop
- R&B, Soul
- Rap
- Reagge
- Rock
- Trap
Training hyperparameters
The following hyperparameters were used during training:
- metric_for_best_model = accuracy
- learning_rate = 5e-5
- seed = 42
- per_device_train_batch_size = 4
- per_device_eval_batch_size = 4
- gradient_accumulation_steps = 1
- warmup_ratio = 0.1
- fp16 = True
- adam_epsilon = 1e-08
- adam_beta1 = 0.9
- adam_beta2 = 0.999